Xiao Liu is a PhD researcher working in the area of Machine Learning. He obtained his undergraduate degree in Mechatronic Engineering from DCU and then completed a master’s in Electronic Engineering at DCU. His main research interests are in the areas of applying data analytics and Machine Learning techniques to metal additive manufacturing and automated optical inspection in the manufacturing process.
Research Interests (Lay Summary)
Xiao Liu started his PhD with I-Form in 2018 and is investigating an adaptive approach to continuous understanding of AM processes through hybrid data streams, particularly on titanium parts. In additive manufacturing of titanium alloys, the formation of defects in parts is typically related to the stability of the meltpool. With increased instability and size of the meltpool comes an increase in the level of emissions generated as the laser processes the material.
The in-situ monitoring system facilitates the collection of large amounts of data during the build process. However, the analysis and characterisation of emissions, and their correlation to defects, is still a manual process that involves looking at the 2D & 3D part representations produced by the monitoring software. Xiao is attempting to explore modern data analytics and Machine Learning techniques to automatically classify huge amounts of data to extract and characterise the most descriptive features.
Given recent advances in computer vision and the availability of potentially large amounts of data collected from the in-situ monitoring system, the idea is to train neural network models on the 2D and 3D representations generated from the additive manufacturing process of titanium alloys to automatically identify the key features of meltpool stability and determine how sensitive certain parameter changes are in the manufacturing process, especially with respect to predicting the presence of defects.
Defects in manufacturing usually result in a waste of time and resources. Ideally, a real time prediction of a defect in the manufacturing process can tell the operator to shut down and stop production at an early stage or intervene whenever possible to prevent further defects from happening. This requires not only the ability to better (and automatically) understand the data, model them and relate them to defects, but it also requires the continuous adjustment of such models as the data is produced.
Additive Manufacturing (3D Printing), Artificial Intelligence, Data Analytics, Predictive Modeling, Real-time Data Analytics